CN105164681B

CN105164681B - The automatic calling of conversational user interface for translation application

Info

Publication number: CN105164681B
Application number: CN201480024228.2A
Authority: CN
Inventors: A·佐梅特; G·恰奇克; M·夏伊纳
Original assignee: Google LLC
Current assignee: Google LLC
Priority date: 2013-03-15
Filing date: 2014-03-14
Publication date: 2018-12-25
Anticipated expiration: 2034-03-14
Also published as: EP2973024A2; CN105164681A; US9195654B2; WO2014143885A3; US20140278346A1; WO2014143885A2

Abstract

A kind of equipment is configured as: receiving translation and inquiry, the translation and inquiry request is by term from source language translation at object language；Determine translation feature associated with the translation and inquiry；Each translation feature into the translation feature assigns characteristic value to form characteristic value；Feature weight is applied to each characteristic value in the characteristic value to generate end value；And determine that provide dialogue translation user interface or non-conversational translates user interface based on whether the end value meets threshold value.Dialogue translation user interface can promote the translation to session, the non-conversational translation user interface can provide translation search as a result, and dialogue translation user interface can be different from non-conversational translation user interface.The equipment is additionally configured to provide the dialogue translation user interface when the end value meets the threshold value for showing.

Description

The automatic calling of conversational user interface for translation application

Background technique

Automatic translation service can be used together with translation application and for providing to such as " by ' what is Your name ' translates into French " or " how with Spanish says ' where is the bus station ' " translation intention Search inquiry answer.Translation and inquiry can appear in several different types of scenes, for example, cypher text Paragraph or the isolated word of another part, translation, translation as sentence of a part talked with etc..Each feelings in these scenes Different translation set can be generated in scape.For example, if user is in foreign country and wants to be conversated with local language, Yong Huke Can need will be isolated sentence from a kind of language (for example, original language or object language) to another kind language (for example, original language or Object language) it translates back and forth.Isolated sentence can create the session of significant problem and/or answer.However, session may It is displayed in equipment in excessively crowded and complicated mode.This for the mobile device with limited display sizes especially such as This.

Summary of the invention

It is described herein to be the purpose for automatically determining translation and inquiry or intention and connect translation and translation user Mouth is suitable for the equipment and technology of the translation and inquiry.If the purpose of translation search inquiry is dialogue, the equipment and skill Art display dialogue translation user interface, the dialogue translate user interface allow users to rapidly and easily with another user Carry out translated session or dialogue.

In some possible embodiments, a kind of method that the one or more processors by equipment execute may include: Translation and inquiry is received, the translation and inquiry request is by one or more terms from source language translation at object language；Determining and institute State the associated one or more translation features of translation and inquiry；Each translation feature into one or more of translation features Assign translation value to form one or more features value；In a manner of linearly or nonlinearly by feature weight be applied to it is one or Each characteristic value in multiple characteristic values is to generate end value；It determines and to provide pair based on whether the end value meets threshold value Words translation user interface or non-conversational translate user interface, and the dialogue translation user interface promotes the translation to session, institute It states non-conversational translation user interface and provides one or more translation searches as a result, and non-conversational translation user interface and institute It is different to state dialogue translation user interface；And the dialogue translation user interface is provided when the end value meets the threshold value For showing.

In some possible embodiments, the method can also include mentioning when the end value is unsatisfactory for the threshold value User interface is translated for showing for the non-conversational.

In some possible embodiments, the method can also include: to receive dialogue training data and non-conversational training Data；It is trained based on one or more of translation features, the dialogue training data and the non-conversational training data Classifier；Based on multiple feature weights are generated to the training of the classifier, it is applied to one or more of spies The feature weight of each characteristic value in value indicative is selected from the multiple feature weight；Based on the multiple feature Weight generates housebroken classifier；Store the housebroken classifier；And it is received from the housebroken classifier It is applied to the feature weight of each characteristic value in one or more of characteristic values.

In some possible embodiments, the method can also include: with associated with the translation and inquiry described Object language generates the translation to the translation and inquiry；It is following for showing to provide to translate user interface via the dialogue Show: with the translation to the translation and inquiry of the object language, and with the translation and inquiry of the original language；With Associated with the translation and inquiry object language generates the potential response to the translation and inquiry；And via described Dialogue translates user interface to provide with the potential response of the object language for showing.

In some possible embodiments, the method can also include: to receive to translate user interface to the dialogue The instruction of user's approval；And the threshold value is changed based on the instruction.

In some possible embodiments, the method can also include: to receive for that will use from dialogue translation The information of family interface exports to the request of message transmission application；And user will be translated from the dialogue based on the request The information of interface exports to the message transmission application.

In some possible embodiments, the method can also include: to translate user interface via the dialogue to mention For the option for changing the original language associated with the translation and inquiry or the object language；And work as the option Change the original language or the object language when being selected by the user.

In some possible embodiments, the method can also include translating user interface via the dialogue to provide With the conversation history of associated with the translation and inquiry original language and the object language.

In some possible embodiments, a kind of computer-readable medium can store one or more instruction, and described one When the one or more processors execution by equipment, so that one or more of processors: reception is turned over for item or a plurality of instruction Inquiry is translated, the translation and inquiry request is by one or more terms from source language translation at object language；The determining and translation Inquire associated one or more translation features；Each translation feature into one or more of translation features is assigned special Value indicative is to form one or more features value；Feature weight is applied to one or more of spies in a manner of linearly or nonlinearly Each characteristic value in value indicative is to generate end value；It determines and dialogue translation to be provided based on whether the end value meets threshold value User interface or non-conversational translate user interface, and the dialogue translation user interface promotes the translation to session, described non-right Words translation user interface provides one or more translation searches as a result, and non-conversational translation user interface and the dialogue It is different to translate user interface；The dialogue translation user interface is provided when the end value meets the threshold value for showing Show；And the non-conversational translation user interface is provided when the end value is unsatisfactory for the threshold value for showing.

In some possible embodiments, the computer-readable medium can also include one or more instruction, described One or more instruction when by one or more of processors execute when so that one or more of processors: with institute The associated object language of translation and inquiry is stated to generate the translation to the translation and inquiry；User is translated via the dialogue Interface is following for showing to provide: with the translation to the translation and inquiry of the object language, and with described The translation and inquiry of original language；It is generated with associated with the translation and inquiry object language to the translation and inquiry Potential response；And via the dialogue translate user interface come provide with the potential response of the object language with In display.

In some possible embodiments, the computer-readable medium can also include one or more instruction, described One or more instruction by one or more of processors when being executed, so that one or more of processors: reception pair Talk about training data and non-conversational training data；Based on one or more of translation features, the dialogue training data and institute Non-conversational training data is stated to train classifier；Multiple feature weights, quilt are generated based on to the training of the classifier The feature weight for each characteristic value being applied in one or more of characteristic values is from the multiple feature weight Selection；Housebroken classifier is generated based on the multiple feature weight；Store the housebroken classifier；And from The housebroken classifier receives the feature power for each characteristic value being applied in one or more of characteristic values Weight.

In some possible embodiments, the computer-readable medium can also include one or more instruction, described One or more instruction by one or more of processors when being executed, so that one or more of processors: being received and is used In the request that the information from the dialogue translation user interface is exported to message transmission application；And based on it is described request come The information from the dialogue translation user interface is exported into the message transmission application.

In some possible embodiments, the computer-readable medium can also include one or more instruction, described One or more instruction by one or more of processors when being executed, so that one or more of processors: being received and is used The instruction of the dialogue translation user interface is ratified at family, and changes the threshold value based on the instruction.

In some possible embodiments, the computer-readable medium can also include one or more instruction, described One or more instruction by one or more of processors when being executed, so that one or more of processors: via institute Dialogue translation user interface is stated to provide for changing the original language associated with the translation and inquiry or the target language The option of speech；And change the original language or the object language when the option is selected by the user.

In some possible embodiments, the computer-readable medium can also include one or more instruction, described One or more instruction by one or more of processors when being executed, so that one or more of processors: via institute Dialogue translation user interface is stated to provide the dialogue of associated with the translation and inquiry original language and the object language History.

In some possible embodiments, a kind of equipment may include one or more processors, one or more of Processor is used for: receiving translation and inquiry, the translation and inquiry request is by one or more terms from source language translation at target language Speech；Determine one or more translation features associated with the translation and inquiry；By feature weight in a manner of linearly or nonlinearly Each translation feature in one or more of translation features is applied to form one or more features weight；Based on described One or more features weights generates end value；It determines and dialogue translation to be provided based on whether the end value meets threshold value User interface or non-conversational translate user interface, and the dialogue translation user interface promotes the translation to session, described non-right Words translation user interface provides one or more translation searches as a result, and non-conversational translation user interface and the dialogue It is different to translate user interface；The dialogue translation user interface is provided when the end value meets the threshold value for showing Show, and provides the non-conversational translation user interface for showing when the end value is unsatisfactory for the threshold value.

In some possible embodiments, one or more of processors be can be also used for: be received described in user's approval The instruction of dialogue translation user interface；And the threshold value is changed based on the instruction.

In some possible embodiments, one or more of processors be can be also used for: be received for that will come from institute The information for stating dialogue translation user interface exports to the request of message transmission application；And it will be from described based on the request The information of dialogue translation user interface exports to the message transmission application.

In some possible embodiments, one or more of processors be can be also used for: receive dialogue training data With non-conversational training data；It is instructed based on one or more of translation features, the dialogue training data and the non-conversational Practice data to train classifier；Multiple feature weights are generated based on to the training of the classifier；Based on the multiple Feature weight generates housebroken classifier；Store the housebroken classifier；And from the housebroken classifier Reception is applied to the feature weight in each of one or more of characteristic values.

In some possible embodiments, one or more of processors be can be also used for: with the translation and inquiry The associated object language generates the translation to the translation and inquiry；User interface is translated via the dialogue to provide Below for showing: with the translation to the translation and inquiry of the object language, and with the institute of the original language State translation and inquiry；The potential sound to the translation and inquiry is generated with the object language associated with the translation and inquiry It answers；And user interface is translated to provide with the potential response of the object language for showing via the dialogue.

In some possible embodiments, a kind of method performed by computer, may include: from user equipment receive with The voice messaging of original language and request for the voice messaging to be translated into object language；It determines for believing with the voice The characteristic value of the associated translation feature of manner of breathing；And it is sent based on the assessment to the characteristic value to the user equipment with institute State object language to the translation of the voice messaging and for described translate to be communicated to the spy of the user of the user equipment Determine user interface or application, the user equipment renders the translation via the user interface or the application.

In some possible embodiments, a kind of system may include: the device for receiving translation and inquiry, the translation Inquiry request is by one or more terms from source language translation at object language；It is associated with the translation and inquiry for determination The device of one or more translation features；Feature is assigned for each translation feature into one or more of translation features Value is to form the device of one or more features value；It is one for being applied to feature weight in a manner of linearly or nonlinearly Or each characteristic value in multiple characteristic values is to generate the device of end value；For based on the end value whether meet threshold value come Determination will provide the device of dialogue translation user interface or non-conversational translation user interface, and the dialogue translation user interface promotees Into the translation to session, the non-conversational translation user interface provides one or more translation searches as a result, and described non-right Words translation user interface is different from dialogue translation user interface；And for being mentioned when the end value meets the threshold value For dialogue translation user interface with the device for display.

Some embodiments of the theme described in the present specification can be carried out to realize one in following advantages A or multiple advantages.User rapidly and easily can carry out translated meeting with another user via dialogue translation user interface Words or dialogue.In some embodiments, the user for accessing foreign country can easily navigate near the foreign country, with institute The resident for stating foreign country conversates.The dialogue translation user interface can be configured as with device display (for example, mobile Device display) boundary in organized mode translated session is provided.

Below with reference to the accompanying drawings the other embodiment of the detailed description disclosure, feature and advantage and the disclosure is each Aspect is structurally and operationally.

Detailed description of the invention

The one or more implementations of drawing illustration that include in the present specification and forming part of this specification Mode and these embodiments are explained together with specification.In the accompanying drawings:

Fig. 1 is the schematic diagram of the general view of example embodiment described herein；

Fig. 2 is the schematic diagram that system and or method described herein may be implemented within example context therein；

Fig. 3 is the stream for creating the instantiation procedure for the housebroken classifier for determining the purpose of translation search inquiry Cheng Tu；

Fig. 4 A and 4B are the exemplary schematic diagrames in conjunction with the process of Fig. 3 description；

Fig. 5 is the purpose for being inquired based on translation search to generate dialogue translation user interface or non-conversational translation user The flow chart of the instantiation procedure of interface；

Fig. 6 A-6G is the exemplary schematic diagram in conjunction with the process of Fig. 5 description；And

Fig. 7 is that the universal computing device that can be used together with technology described herein and General Mobile calculate equipment Exemplary schematic diagram.

Specific embodiment

It is described in detail below to refer to attached drawing.Same reference mark in different figures can identify the same or similar member Part.

System and or method described herein can automatically determine the purpose or intention of translation search inquiry, and can So that translation and translation user interface are suitable for the translation search inquiry.For example, if the purpose of translation search inquiry It is dialogue, then the system and or method can provide dialogue translation user interface for showing.User is translated in the dialogue Interface can promote the translation to the session in the boundary of device display.If the purpose of the translation search inquiry is pair Other purposes other than words, then the system and or method can provide non-conversational translation user interface for showing.It is described Non-conversational translation user interface can provide the link for example to the website with translation information.

Fig. 1 is the schematic diagram of the general view of example embodiment described herein.For general view, it is assumed that equipment is not (in Fig. 1 Shown in) associated with the first user (user 1), the first user wants to carry out translated meeting with second user (user 2) Words.It is also assumed that the first user says English and assumes that second user is spoken French.The equipment can draw with search is provided to equipment Hold up server apparatus (not shown in FIG. 1) communication of user interface.First user can use the equipment and come to search Engine owners interface provides translation and inquiry (for example, " what is your name " is translated into French).The equipment can be to Server apparatus provides translation and inquiry.

Server apparatus can receive translation and inquiry, and can determine one or more features in translation and inquiry.? In some embodiments, feature may include appearance of the specific word in translation and inquiry, the word in translation and inquiry quantity, with turn over Translate the associated original language of inquiry and/or object language etc..Server apparatus can assign characteristic value to identified feature, and And feature weight can be applied to corresponding characteristic value to generate end value in a linear or non-linear fashion.Server apparatus Whether can meet threshold value based on the end value to determine and to provide dialogue translation user interface or non-conversational translation user Interface.Dialogue translation user interface can promote the translation to session, and talk with translation user interface can be turned over non-conversational Translate user interface difference.

Server apparatus can also provide dialogue translation user interface to equipment when end value meets threshold value.As shown in figure 1 It shows, the equipment can provide dialogue translation user interface for showing to the first user.Dialogue translation user interface It can show the english language version and French version of phrase " what is your name ", and whether can inquire the first user Approval dialogue translation user interface.If the first user indicates approval dialogue translation user interface, server apparatus can benefit Conversational user interface is made more often to be displayed to the first user to change the threshold value with the approval.

As further illustrated in Figure 1, second user can provide the response to the equipment (for example, " mon nom Est Paul "), and the equipment can receive the response.The equipment can provide the response to server apparatus, And server apparatus can be translated (for example, translating into " My name is Paul ") to the response.Server apparatus Dialogue translation user interface can be updated to include the response, and updated dialogue translation can be provided to the equipment User interface is for showing.As further illustrated in Figure 1, the equipment can be to the first user and/or second user Updated dialogue translation user interface is provided for showing.Updated dialogue translation user interface can provide object (for example, button, icon, link etc.), the object can make the equipment translate updated dialogue when selected User interface exports to message transmission application, e-mail applications, chat application etc..

Such arrangement can enable a user to via dialogue translation user interface rapidly and easily with another user Carry out translated session or dialogue.In some embodiments, this can enable the users of access foreign country can be described outer State nearby navigates, and conversates with the resident of the foreign country.Dialogue translation user interface can be configured as with equipment Organized mode in the boundary of display (for example, mobile device display) provides translated session.

User interface as described in this article may include graphical user interface (GUI) and/or non-Graphic User's Interface, Such as text based interface.User interface can be via custom interface (for example, proprietary interface) and/or other kinds of interface (for example, interface etc. based on browser) provides a user information.In some embodiments, user interface can be via one Or multiple input equipments receive users' input, can be that user can configure (for example, user can change the ruler of user interface Information that is very little, being shown in user interface, the Color scheme used by user interface, text, image, icon, window etc. with Position, etc. in the interface of family).In some embodiments, user interface can not be what user can configure.With user interface Associated information can be selected by user's (for example, via touch screen, mouse, keyboard, keypad, voice command etc.) of equipment And/or manipulation.

It is storable that term document as used in this article should be broadly interpreted to include any machine readable and machine Achievement.Document may include for example Email, file, file combination, with arrive alternative document inline linking One or more files, newsgroup posts, blog, the web advertisement, uniform resource locator (URL) etc..In the upper of internet Hereinafter, common document is webpage.Webpage generally include text information and may include embedded information (such as metamessage, Image, hyperlink etc.) and/or embedded instruction (for example, Javascript etc.).

Fig. 2 is the schematic diagram that system and or method described herein may be implemented within example context 200 therein. As shown, environment 200 may include via network 230 and the client device 210 interconnected of server apparatus 220.Ring The component in border 200 can be interconnected via the combination of wired connection, wireless connection or wired connection and wireless connection.

Client device 210 may include one or more calculating equipment, for example laptop computer, individual Computer, tablet computer, desktop computer, workstation computer, smart phone, personal digital assistant (PDA) and/or other Calculate equipment.In some embodiments, client device 210 may include browser application, the browser application with Search engine 225 providing in server apparatus 220 and described below communicates.

Server apparatus 220 may include one or more server apparatus or other kinds of calculating and communication equipment. Server apparatus 220 may include the equipment that can be communicated with client device 210 (for example, via network 230).One In a little embodiments, server apparatus 220 may include one or more laptop computers, personal computer, work station meter Calculation machine, server, central processing unit (CPU), graphics processing unit (GPU), specific integrated circuit (ASIC), field-programmable Gate array (FPGA) etc..In some embodiments, server apparatus 220 may include search engine 225 and can execute Some or all of the function of being described herein for client device 210 function.Alternatively, server apparatus 220 can be with Be omitted and client device 210 can execute it is complete in function that is described herein, such as being executed by server apparatus 220 Portion's function.

Search engine 225 can enable server apparatus 220 to receive the search inquiry from client device 210, And document is searched in response to described search inquiry.Search engine 225 can enable server apparatus 220 to be based on to text The described search of shelves provides search result to client device 210 to generate search result.In some embodiments, Search result may include the result row of referred to as search-engine results page.In some embodiments, search engine 225 can be with The user of client device 210 is enabled to provide translation search inquiry to server apparatus 220.As described in this article, it takes Business device equipment 220 can receive translation search and inquire and can provide a user dialogue translation user interface or non-conversational translation User interface.

Network 230 may include the network of such as local area network (LAN), wide area network (WAN), Metropolitan Area Network (MAN) (MAN), such as public The telephone network of switched telephone network (PSTN) or cellular network, Intranet, the combination of internet or network.

While figure 2 show that the example apparatus of environment 200, but in some embodiments, environment 200 may include with The equipment described in Fig. 2 compares less equipment, different equipment, the equipment being arranged differently or additional equipment.It is additional Ground or alternatively, one or more equipment of environment 200 can execute the one or more other equipment being described as by environment 200 One or more tasks of execution.

Fig. 3 is for creating the instantiation procedure 300 for the housebroken classifier for determining the purpose of translation search inquiry Flow chart.In some embodiments, process 300 can be executed by server apparatus 220.In some embodiments, mistake Journey 300 can be by separating or another equipment or equipment group including server apparatus 220 is (for example, client with server apparatus 220 End equipment 210) it executes.

As shown in Figure 3, process 300 may include receiving information (frame 310) associated with translation feature.For example, Server apparatus 220 can provide the use for allowing users to the feature (or standard) of purpose of definition instruction translation search inquiry Family interface is for showing.In some embodiments, feature can be automatically generated by server apparatus 220 rather than be determined by user Justice.In some embodiments, user can mark search inquiry based on how user classifies to search inquiry.Example Such as, search inquiry can be labeled as with " dialogue is intended to " or have " non-conversational intention " by user.

In some embodiments, feature may include appearance of the specific word in translation search inquiry.For example, such as The term of courtesy of the interrogative of " what " or " how " and such as " please " and " Thank you " can be under the backgrounds of dialogue More frequently.In some embodiments, feature may include the quantity of the word in translation search inquiry.For example, as translation is searched The quantity of word in rope inquiry increases, and the purpose of translation search inquiry is that the probability of dialogue increases.In some embodiments, special Sign may include original language associated with translation search inquiry and/or object language.For example it is assumed that the source of translation search inquiry Language or object language are Latin languages.This can indicate that the purpose of translation search inquiry is not dialogue (for example, being non-conversational), because For Latin language be rarely employed or dead language.

In some embodiments, feature may include geographical location associated with client device 210 and and client The associated geographical location of the user of end equipment 210.For example it is assumed that geographical location associated with the user be the U.S. and with visitor The associated geographical location of family end equipment 210 is non-English speaking country (for example, France).This can indicate user be on a journey and The purpose of translation search inquiry is dialogue.In some embodiments, feature may include that contamination or syntactic structure are being turned over Translate the appearance in search inquiry.For example it is assumed that translation search inquiry includes contamination, such as " Is this way to XXX? " or " I am lost ".The purpose that such contamination can indicate translation search inquiry is dialogue.

In some embodiments, feature may include and provide the equipment of translation search inquiry (for example, client device 210) associated type.For example, the translation search inquiry provided by mobile device (for example, smart phone) can indicate to translate The purpose of search inquiry is dialogue, and the translation search inquiry provided by fixed equipment (for example, desktop computer) can indicate The purpose of translation search inquiry is not dialogue.In some embodiments, feature may include the user with client device 210 Associated information (for example, gender, age, work title etc.).For example it is assumed that the user of client device 210 is continually Go abroad the sales force of travelling.This can indicate that the purpose of translation search inquiry is dialogue.In some embodiments, it services Device equipment 220 can store identified feature.

As further shown in Fig. 3, process 300 may include receiving dialogue training data and non-conversational training data (frame 320).For example, server apparatus 220, which can provide, allows users to mark dialogue training data and non-conversational training number According to position (for example, filename) user interface with for showing.In some embodiments, server apparatus 220 can be with Dialogue training data and non-conversational training data are created according to the information retrieved from each provenance.For example, server apparatus 220 It can be according to the anonymity for example received from service provider (for example, Internet Service Provider, telco service provider etc.) Chat conversations writing record, telephone conversation writing record etc. create dialogue training data.Server apparatus 220 can according to by The random search that search engine 225 receives is inquired to create non-conversational training data.

In some embodiments, server apparatus 220 can be from such as robot of Turkey (Mechanical Turk) Source receive dialogue training data and non-conversational training data.Robot of Turkey may include enabling computer programmer Coordinate the use of human intelligence to execute the crowdsourcing Internet market for the task that computer can not currently carry out.For example, native ear Its robot can provide text information to human work person, and text information can be labeled as and dialogue by human work person Or non-conversational is related.It is high probability associated with dialogue (for example, greater than threshold value that dialogue training data, which may include that mark has, ) text information.Non-conversational training data may include mark have be not high probability associated with dialogue (for example, Less than or equal to threshold value) text information.In some embodiments, server apparatus 220 can store dialogue training number According to non-conversational training data.

As further shown in Fig. 3, process 300 may include being trained based on translation feature information and training data Classifier (frame 330).For example, server apparatus 220 can be based on information associated with feature, dialogue training data and non- Talk with training data to train classifier.In some embodiments, classifier may include Machine learning classifiers, such as base Programmed in support vector machines, boosting, decision tree learning, correlation rule study, neural network, genetic planning, inductive logic, The classifier of cluster, Bayesian network, intensified learning, expression study, similarity measurement study, sparse dictionary study etc..One In a little embodiments, classifier may include the classifier based on random regularization on-line study.

In some embodiments, server apparatus 220 can be by belonging to specified classification (for example, dialogue known to offer Training data and non-conversational training data) information and classifier created based on the Given information train classifier.Training point Class device can be iterative process.

Fig. 3 is returned to, process 300 may include the feature weight generated based on the training to classifier for translation feature (frame 340).For example, server apparatus 220 can be classified based on being trained using dialogue training data and non-conversational training data Device and generate the feature weight for feature.In some embodiments, server apparatus 220 can be based on special characteristic right There are how many times in words training data or non-conversational training data to calculate the feature weight for special characteristic.For example it is assumed that Determining feature (the A associated with appearance of the specific word in translation search inquiry of server apparatus 220₁) appear in dialogue training 30 (30) are secondary in data.It is further assumed that dialogue training data includes 100 times (100) observations.In such a scenario, it services Device equipment 220 can will be directed to feature (A₁) feature weight (w_A1) it is calculated as " 0.3 ".

In some embodiments, feature weight can be any value between " 0 " and " 1 " and can depend on The associated feature of feature weight.For example, the feature of more preferable instruction dialogue translation can be given and give less instruction dialogue The feature weight of the feature of translation compares higher feature weight.

As further shown in Fig. 3, process 300 may include that housebroken classifier is generated based on feature weight (frame 350).For example, server apparatus 220 can use feature weight to generate housebroken classifier.In some embodiments In, server apparatus 220 can generate housebroken classifier according to following equation:

End value=w_A1×A₁+w_B1×B₁+w_C1×C₁+w_D2×D₂+w_E3×E₃...,

Wherein, end value may include the final calculating inquired for translation search by housebroken classifier and generated, w_A1It may include fisrt feature (A₁) feature weight, w_B1It may include second feature (B₁) feature weight, etc..Some In embodiment, each feature can be referred to by alphabetical (for example, A, B, C etc.), and can be by number for the type of feature (for example, 1,2,3 etc.) refer to.For example, feature associated with appearance of the specific word in translation search inquiry can use word Female A is referred to.The interrogative (for example, " what ") of feature (A) can use digital one (1) to refer to, so that word " what " can To utilize term A₁To refer to.The greeting term (for example, " Thank you ") of feature (A) can use digital two (2) to refer to Generation, so that word " Thank you " can use term A₂To refer to.

It in some embodiments, can be based on any non-linear of feature by the end value of housebroken classifier calculated Function, for example polynomial function, the set of logical condition, combination of nonlinear function etc..In some embodiments In, end value, which calculates, can be based on machine learning method, such as support vector machines, boosting, decision tree learning, correlation rule Study, genetic planning, inductive logic programming, cluster, Bayesian network, intensified learning, indicates study, similitude at neural network Metric learning, sparse dictionary study etc..Housebroken classifier can not only provide weight for single feature but also can To provide weight for the nonlinear combination of feature.

Fig. 3 is returned to, process 300 may include storing housebroken classifier (frame 360).For example, server apparatus 220 can Being stored in housebroken classifier in memory devices associated with server apparatus 220.In some embodiments, Server apparatus 220 can provide housebroken to search engine 225 and/or by the translation application that server apparatus 220 provides Classifier.As described below, search engine 225 and/or translation application can use housebroken classifier to determine translation The purpose of search inquiry is to talk with or non-conversational.

Although Fig. 3 process 300 is shown as include specific quantity and arrangement frame, in some embodiments, mistake Journey 300 may include the different arrangements of less frame, additional frame or frame.Additionally or alternatively, frame can be performed in parallel In some frames.

Fig. 4 A and 4B are the schematic diagrames in conjunction with the example 400 of the process of Fig. 3 description.In example 400, it is assumed that server is set Standby 220 provide the user interface 410 for allowing users to create housebroken classifier for showing.As shown, user Interface 410 can request user to provide translation feature information.It is further assumed that user defines above combination Fig. 3 description, user Feature in interface 410.User can also define type associated with the feature in user interface 410.For example, user can be with Definition is directed to the feature (A of interrogative " what "₁) and can define for the feature (A for greeting term " Thank you "₂).With Family can define the feature (B for phrase " Is this way to XXX "₁) and can define for word combination " I Feature (the B of am lost "₂)。

In some embodiments, feature can be predefined by server apparatus 220 rather than is defined by the user.In some realities It applies in mode, user can mark search inquiry based on how user classifies to search inquiry.For example, can require to use Search inquiry is labeled as with " dialogue is intended to " or has " non-conversational intention " by family.

As shown in Figure 4 A further, user interface 410 can enable a user to mark dialogue training data and The position (for example, filename) of non-conversational training data.In example 400, it is assumed that user provide include dialogue training data and The position (for example, filename) of the file of non-conversational training data.Server apparatus 220 can talk with training number from document retrieval According to non-conversational training data, and can store dialogue training data and non-conversational training data.

In some embodiments, server apparatus 220 can create dialogue according to the information retrieved from each provenance Training data and non-conversational training data.For example, server apparatus 220 can be from anonymous chat writing record, telephone conversation text Word record etc. creates dialogue training data.Server apparatus 220 can from the search inquiry received by search engine 225 Lai Create non-conversational training data.In some embodiments, server apparatus 220 can connect from the source of such as robot of Turkey Receive dialogue training data and non-conversational training data.

As shown in Figure 4 A further, is defining the feature for classifier and receiving dialogue training data After non-conversational training data, user can choose object (for example, creation classifier button, icon, link etc.).To object Selection server apparatus 220 can be made to train based on the feature of definition, dialogue training data and non-conversational training data Classifier.Server apparatus 220 can be given birth to based on classifier is trained using dialogue training data and non-conversational training data At the feature weight for feature.Server apparatus 220 can generate housebroken classifier based on feature weight and can To store housebroken classifier.In some embodiments, server apparatus 220 can be to search engine 225 and/or by taking The translation application that device equipment 220 of being engaged in provides provides housebroken classifier.

As shown in Fig. 4 B, after generating and storing housebroken classifier, server apparatus 220 can be provided Indicate the user interface 420 for having generated housebroken classifier for showing.As further shown, user interface 420 can request user's setting for the threshold value of dialogue translation.It can be used by server apparatus 220 for the threshold value of dialogue translation In in response to translation search inquiry come determine to provide dialogue translation user interface or non-conversational translate user interface to be used for Display.The further details of the threshold value for dialogue translation are provided below.In example 400, such as further show in figure 4b , it is assumed that user will be set as particular value (for example, 0.75) for the threshold value of dialogue translation first.

Fig. 5 is the purpose for being inquired based on translation search to generate dialogue translation user interface or non-conversational translation user The flow chart of the instantiation procedure 500 of interface.In some embodiments, process 500 can be executed by server apparatus 220.? In some embodiments, process 500 can be by separating or another equipment including server apparatus 220 with server apparatus 220 (for example, client device 210) or equipment group execute.

As shown in Figure 5, process 500 may include receiving translation search inquiry (frame 510).For example, being set with client Standby 210 associated users can provide translation search inquiry to client device 210.In some embodiments, Yong Huke To provide translation search inquiry to the browser application or translation application executed on client device 210.In some embodiment party In formula, user can come to input translation search inquiry to client 210 via the mechanism of keyboard, touch-screen display etc.. In some embodiments, user can provide translation search inquiry to client device 210 by word of mouth.

Translation search inquiry can be in order to talk with purpose or for non-conversational purpose.For example, such as " by what is Your name translates into French " translation search inquiry can be to talk with purpose.Such as " car is translated into Spain The translation search inquiry of language " can be for non-conversational purpose.In some embodiments, from 210 (example of mobile client device Such as, smart phone) the translation search inquiry that receives can be to talk with purpose, and from 210 (example of stationary client devices Such as, desktop computer) receive translation search inquiry can be for non-conversational purpose.

Client device 210 can provide translation search inquiry to server apparatus 220.Server apparatus 220 can connect Translation search inquiry is received, and in some embodiments, can store translation search inquiry.

As being also shown in FIG. 5, process 500 can include determining that one or more features in translation search inquiry (frame 520).For example, server apparatus 220 can determine whether translation search inquiry includes in the above feature defined in conjunction with Fig. 3 Any feature.In some embodiments, server apparatus 220 can serve as interpreter search inquiry include specific word (for example, doubt Ask word, term of courtesy etc.) or when certain amount of word (for example, be greater than three words, four words etc.) determine translation search inquiry packet Include feature.Server apparatus 220 can serve as interpreter the original language of search inquiry or object language includes the language (example being rarely employed Such as, Latin language) when determine translation search inquiry include feature.Server apparatus 220 can serve as interpreter search inquiry include word or Syntactic structure specific combination (for example, " where is airport? ") when determine translation search inquiry include feature.

In some embodiments, server apparatus 220 can attribute based on client device 210 and/or and client The associated user of end equipment 210 come determine translation search inquiry include feature.For example, server apparatus 220 can work as client The geographical location of end equipment 210 determines that translation search inquiry includes feature when being in the foreign country of the geographical institute of living of user.Clothes Business device equipment 220 when client device 210 can be mobile device and when non-stationary device determines that translation search inquiry includes spy Sign.Server apparatus 220 can be on a journey as user property instruction user (for example, the calendar of user can indicate that user exists Specific date goes to German travelling) when determine translation search inquiry include feature.

As further shown in Fig. 5, process 500 may include assigning characteristic value (frame 530) to identified feature. For example, server apparatus 220 can assign characteristic value to each feature for being determined as being present in translation search inquiry.Some In embodiment, server apparatus 220 can assign particular characteristic value (example to each feature being present in translation search inquiry Such as, " 1 "), and can be assigned to each feature being not present in translation search inquiry another particular characteristic value (for example, "0").Such arrangement may insure that the feature being present in translation search inquiry is given weight appropriate, and be not present Feature in translation search inquiry is not given any weight.In some embodiments, server apparatus 220 can be to depositing It is that the different characteristic in translation search inquiry assigns different characteristic values (for example, to fisrt feature assigned value " 1 ", to second Feature assigned value " 3 ", etc.).

In some embodiments, characteristic value can be any value between " 0 " and " 1 " and can depend on and spy The associated feature of value indicative.For example, the feature of more preferable instruction dialogue translation can be given and give less instruction dialogue translation Feature characteristic value compare higher characteristic value.

In some embodiments, server apparatus 220 can provide the reference to characteristic value.For example, server apparatus 220 can will refer to (A₁) it is supplied to the fisrt feature of the first kind, it will refer to (A₂) it is supplied to the fisrt feature of Second Type, It will refer to (B₁) it is supplied to third feature of the first kind, etc..

Fig. 5 is returned to, process 500 may include that the feature weight from housebroken classifier is applied to corresponding feature Value is to generate end value (frame 540).For example, server apparatus 220 can be by above in conjunction with the housebroken of Fig. 3-Fig. 4 B description The feature weight of classifier is applied to corresponding characteristic value, and can calculate end value based on feature weight and characteristic value. In some embodiments, server apparatus 220 can be according to following weighted sum equation come by the spy of housebroken classifier Sign weight is applied to characteristic value:

End value=w_A1×A₁+w_B1×B₁+w_C1×C₁+w_D2×D₂+w_E3×E₃...。

As further shown in Fig. 5, process 500 can include determining that whether end value meets threshold value (frame 550).Example Such as, server apparatus 220 can determine whether end value meets the threshold value for dialogue translation.Threshold value can be by server apparatus 220 in response to translation search inquiry come determine to provide dialogue translation user interface or non-conversational translate user interface with For showing.In some embodiments, it can satisfy threshold value when end value is greater than threshold value, and when end value is less than or waits It is unsatisfactory for threshold value when threshold value, or vice versa.

Fig. 5 is returned to, if end value is unsatisfactory for threshold value (frame 550- is no), process 500 may include generating non-conversational to turn over Translate user interface (frame 560).For example, if server apparatus 220 determines that end value is unsatisfactory for threshold value, server apparatus 220 Non-conversational translation user interface can be generated.In some embodiments, non-conversational translation user interface may include by searching for The user interface and/or the translation application provided by server apparatus 220 that engine 225 generates.In some embodiments, described It is different from dialogue translation user interface that non-conversational translates user interface.Non-conversational translation user will be described with reference to Fig. 6 D to connect The example of mouth.

As further shown in Fig. 5, process 500 may include providing non-conversational translation user interface for showing (frame 570).For example, server apparatus 220 can provide non-conversational translation user interface generated to client device 210. Client device 210 can receive non-conversational translation user interface, and can show that non-conversational translates user interface to user. User can translate user interface with non-conversational and interact for example to select linking for the website for providing translation information.

Fig. 5 is returned to, if end value meets threshold value (frame 550- is), process 500 may include generating dialogue translation to use Family interface (frame 580).For example, server apparatus 220 can give birth to if server apparatus 220 determines that end value meets threshold value User interface is translated at dialogue.In some embodiments, dialogue translation user interface can be configured as with device display Organized mode in the boundary of (for example, mobile device display) provides translated session.Fig. 6 E and 6G will be referred to To describe the example of dialogue translation user interface.

As further shown in Fig. 5, process 500 may include providing dialogue translation user interface for showing (frame 590).For example, server apparatus 220 can provide dialogue translation user interface generated to client device 210.Visitor Family end equipment 210 can receive dialogue translation user interface, and dialogue translation user interface can be shown to user.

Although Fig. 5 process 500 is shown as include specific quantity and arrangement frame, in some embodiments, mistake Journey 500 may include the different arrangements of less frame, additional frame or frame.Additionally or alternatively, frame can be performed in parallel In some frames.

Fig. 6 A-6G is the schematic diagram in conjunction with the example 600 of the process of Fig. 5 description.In example 600, it is assumed that user utilizes Client device 210 is come the translation application that accesses search engine 225 or provided by server apparatus 220.As shown in Fig. 6 A, When client device 210 accesses search engine 225 or translation application, server apparatus 220 can be mentioned to client device 210 For user interface 610, and client device 210 can provide a user user interface 610 for showing.User interface 610 may include search box, and user can provide translation search inquiry 620 in described search frame.In some embodiments, User can input translation search inquiry 620 by word of mouth to client device 210.As shown in fig. 6 further, it is assumed that User inputs " " what is your name " is translated into French " as translation search inquiry 620.

Client device 210 can provide translation search inquiry 620, and server apparatus 220 to server apparatus 220 It can receive translation search inquiry 620.Server apparatus 220 can determine whether translation search inquiry 620 includes above combine Any feature in the feature that Fig. 3 is defined.For example, as shown in Fig. 6 B, server apparatus 220 can determine word " what " with 630 (the A of fisrt feature of the first kind₁) associated, and the 630 (B of second feature of word " is " and the first kind₁) associated.Clothes Business device equipment 220 can determine the 630 (C of third feature of word " your " Yu the first kind₁) associated, and word " name " and the 630 (the D of fourth feature of two types₂) associated.Server apparatus 220 can determine the 5th spy of word " French " and third type Levy 630 (E₃) associated.

As shown in fig. 6b further, server apparatus 220 can be to being determined as being present in translation search inquiry 620 In each feature 630 assign characteristic value 640.In some embodiments, server apparatus 220 can be searched to translation is present in Each feature 630 in rope inquiry 620 assigns particular characteristic value 640 (for example, " 1 ").For example, server apparatus 220 can be to Feature 630 is (for example, A₁、B₁、C₁、D₂And E₃) assign particular characteristic value 640.In example 600, server apparatus 220 can be with Characteristic value 640 is referred to as A₁、B₁、C₁、D₂And E₃.Server apparatus 220 can be inquired in 620 to translation search is not present in Each feature assigns another particular characteristic value 640 (for example, " 0 ").

As shown in Fig. 6 C, server apparatus 220 can be by the above housebroken classification in conjunction with Fig. 3-Fig. 4 B description The feature weight 650 of device is applied to corresponding characteristic value 640, and can be calculated based on feature weight 650 and characteristic value 640 End value.As further shown, server apparatus 220 can answer feature weight 650 according to following weighted sum equation Use characteristic value 640:

End value=w_A1×A₁+w_B1×B₁+w_C1×C₁+w_D2×D₂+w_E3×E₃...

End value may include the final calculating generated by housebroken classifier for translation search inquiry 620, w_A1It can To include 640 (A of characteristic value₁) feature weight 650, w_B1It may include 640 (B of characteristic value₁) feature weight 650, w_C1It can be with Feature weight 650, w including characteristic value 640 (C1)_D2It may include 640 (D of characteristic value₂) feature weight 650, and w_E3It can To include 640 (E of characteristic value₃) feature weight 650.

Server apparatus 220 can determine whether end value meets the threshold value for dialogue translation.If server apparatus 220 determine that end value is unsatisfactory for threshold value, then non-conversational translation user interface 660 can be generated in server apparatus 220.Server is set Standby 220, which can provide non-conversational to client device 210, translates user interface 660.As shown in Fig. 6 D, client device 210 can receive non-conversational translation user interface 660, and can show that non-conversational translates user interface 660 to user.It is non-right Words translation user interface 660 may include the user interface generated by search engine 225 and/or be provided by server apparatus 220 Translation application.As shown in also in figure 6d, non-conversational translates user interface 660 can be by word " hello " turning over to French It translates in the frame that (for example, " bonjour ") is shown in below search input frame, and can show that can provide word " hello " arrives The ranked list of the document of the translation of French.In this case, server apparatus 220 can determine that translation search is inquired " hello is translated into French " be not associated with dialogue.

If server apparatus 220 determines that end value meets threshold value, server apparatus 220 can be generated dialogue translation and use Family interface 670.Server apparatus 220 can provide dialogue translation user interface 670 to client device 210.As shown in Fig. 6 E Out, client device 210 can receive dialogue translation user interface 670, and dialogue translation user can be shown to user Interface 670.Dialogue translation user interface 670, which can be configured as in a manner of organized in the boundary of device display, to be mentioned For translated session.For example, dialogue translation user interface 670 can be talked with French and English to show, can help (for example, English-speaking) user conversates with the people to speak French.In some embodiments, user can change dialogue translation The original language (for example, English) and/or object language (for example, French) of user interface 670.For example, user can exchange dialogue Translate the source language and the target language of user interface 670.

In some embodiments, dialogue translation user interface 670 can be shown with both source language and the target language Whole history of dialogue or session.As further shown in Fig. 6 E, dialogue translation user interface 670 can request user to refer to Show whether user likes dialogue translation user interface 670.If a user indicate that user likes dialogue translation user interface 670, then Server apparatus 220 is adjustable for the threshold value for talking with translation, so that dialogue translation user interface 670 is more frequently for use Family occurs.For example, server apparatus 220 can reduce threshold value when user indicates that user likes dialogue translation user interface 670. If a user indicate that user does not like dialogue translation user interface 670, then server apparatus 220 is adjustable for dialogue translation Threshold value so that dialogue translation user interface 670 less frequently for user occur.

As further shown in Fig. 6 E, dialogue translation user interface 670 can show object (for example, export button, Icon, link etc.), the object can be wanted dialogue translation user interface 670 exporting to message transmission using (example as user Such as, e-mail applications, text message applications etc.) when selected by user.When user's selecting object, server apparatus 220 can To indicate that client device 210 provides dialogue translation user interface 670 to the application of the message transmission of client device 210.For example, As shown in Fig. 6 F, client device 210 can automatically open Email User interface 680, and can use and come from The information of dialogue translation user interface 670 fills the text of email message automatically.User can be provided for electronics postal The addressee of part message, and can indicate that client device 210 sends email message to addressee.

In some embodiments, dialogue translation user interface 670 can show before providing response and dive to problem It is responding.For example, dialogue translation user interface 670 can be with English (example as indicated by the reference marker 690 in Fig. 6 G Such as, " What is your name? ") and display problem come with French (for example, " quel est votre nom ").Such as into one Shown in step, dialogue translation user interface 670 can show the potential response to problem (for example, " mon nom with French est XXX”)。

In some embodiments, dialogue translation user interface 670 may be used to provide to be directed to and be broadcast by client device 210 The translated subtitle of video put or that (for example, via camera) is captured by client device 210.For example, if user passes through French video is accessed by client device 210, then server apparatus 220 can provide dialogue translation to client device 210 and use Family interface 670 allows to provide English subtitles to French video.

In some embodiments, if server apparatus 220 detects that client device 210 is going into foreign country, Server apparatus 220 can use dialogue translation user interface 670 and carry out automatic translation by the addressable text of client device 210 Shelves.For example, if user travels in transit to Spain and needs to show English documents, server apparatus to Spain client File can be automatically translated into western class when server apparatus 220 detects that client device 210 is going into Spain by 220 Tooth language.In some embodiments, server apparatus 220 can detect client device 210 when server apparatus 220 Spain's phrase document or handbook are downloaded into client device 210 when into Spain.

Fig. 7 is that the universal computing device 700 that can be used together with technology described herein and General Mobile calculating are set Standby 750 exemplary schematic diagram.Universal computing device 700 or General Mobile, which calculate equipment 750, can correspond to such as client Equipment 210 and/or server apparatus 220.It calculates equipment 700 and is intended to indicate that various forms of digital computers, it is such as on knee Computer, desktop computer, work station, personal digital assistant, server, blade server, host and other suitable calculating Machine.Mobile computing device 750 is intended to indicate that various forms of mobile devices, such as personal digital assistant, cellular phone, intelligence electricity Words, tablet computer and other similar calculating equipment.Component shown in Fig. 7, its connection and relationship and its function are intended to only It is merely illustrative, and it is not intended to be limited to embodiment described herein.

Calculating equipment 700 may include processor 702, memory 704, storage equipment 706, is connected to 704 and of memory The high-speed interface 708 of high-speed expansion ports 710 and the low-speed interface for being connected to low-speed expansion port 714 and storage equipment 706 712.Each component in component 702,704,706,708,710,712 and 714 is connected with each other using various buses, and can be with It is installed on common mainboard or otherwise installs in appropriate circumstances.Processor 702 is capable of handling by based on The instruction executed in calculation equipment 700 is used for including storage in memory 704 or in storage equipment 706 to be such as coupled to Graphical user interface (GUI) on the external input/output device of the display 716 of high-speed interface 708 shows graphical information Instruction.In other embodiments, multiple processors and/or multiple buses can be used in appropriate circumstances, together with multiple Memory and a plurality of types of memories.In addition, multiple equipment 700 that calculate can be with each of the part of the necessary operation of offer Equipment is connected to become server library, blade server group or multicomputer system etc..

Memory 704 stores the information calculated in equipment 700.In one embodiment, memory 704 include one or Multiple volatile memory-elements.In another embodiment, memory 704 includes one or more nonvolatile memory lists Member.Memory 704 can also be another form of computer-readable medium, such as disk or CD.Computer-readable medium can To refer to non-transient memorizer equipment.Memory devices can refer to the single memory space stored in equipment or across multiple storages The memory space of device distribution.

Storage equipment 706 can provide massive store to calculate equipment 700.In one embodiment, equipment is stored 706 can be or comprising computer-readable medium, such as floppy device, hard disc apparatus, compact disk equipment or tape unit, flash memory Reservoir or other kinds of solid-state memory device or equipment array, including setting in storage area network or other configurations It is standby.Computer program product can be tangibly embodied in the information carrier.Computer program product can also include instruction, institute State one or more methods that instruction executes such as those described herein upon being performed.Information carrier is computer or machine Memory on device readable medium, such as memory 704, storage equipment 706 or processor 702.

The management of high-speed interface 708 is for the bandwidth-intensive operations for calculating equipment 700, and the management of low-speed interface 712 is compared with low strap Wide intensive action.Such distribution of function is only example.In one embodiment, high-speed interface 708, which is coupled to, deposits Reservoir 704, display 716 (for example, passing through graphics processor or accelerator), and high-speed expansion ports 710 are coupled to, It can receive various expansion cards.In this embodiment, low-speed interface 712 can be coupled to storage equipment 706 and low speed expands Open up port 714.It may include the low-speed expansion of various communication port (for example, USB, bluetooth, Ethernet, wireless ethernet etc.) Port 714 can be coupled to one or more input-output apparatus, such as keyboard, pointing device, scanner or for example pass through The equipment of network adapter networking, such as switch or router.

As shown in figures, calculating equipment 700 can implement in many different forms.For example, calculating equipment 700 It may be implemented as standard server 720, or be carried out repeatedly in such server group.Calculating equipment 700 can also quilt It is embodied as a part of frame server system 724.In addition, calculating equipment 700 may be implemented within such as laptop computer In 722 personal computer.It alternatively, can be with the movement of such as mobile computing device 750 from the component for calculating equipment 700 Other component in equipment (not shown) is combined.Each equipment in such equipment may include calculate equipment 700, One or more in 750 calculate equipment, and whole system can be by communicate with one another 700,750 groups of multiple calculating equipment At.

Mobile computing device 750 may include the input/output of processor 752, memory 764, such as display 754 (I/O) equipment, communication interface 766 and transceiver 768 and other component.Mobile computing device 750 may be provided with having There is the storage equipment of such as mini drive or other equipment to provide additional storage.Component 752,764,754,766 and 768 In each component be connected with each other using various buses, and if the dry part in component can be installed on common mainboard or It otherwise installs in appropriate circumstances.

Processor 752 is able to carry out the instruction in mobile computing device 750, including the instruction being stored in memory 764. Processor 752 may be implemented as include separation multiple analog processors and digital processing unit chip chipset.Processing Device 752 can provide the coordination of the other component for example to mobile computing device 750, such as to user interface, by mobile computing The control of application and the wireless communication by mobile computing device 750 that equipment 750 is run.

Processor 752 can be communicated by the control interface 758 and display interface 756 for being coupled to display 754 with user. Display 754 can be, such as Thin Film Transistor-LCD (TFT LCD) or Organic Light Emitting Diode (OLED) display Or other display technologies appropriate.Display interface 756 may include for driving display 754 so that graphical information is presented to user With the circuit appropriate of other information.Control interface 758 can receive order from the user and conversion command for mentioning Give processor 752.It is communicated in addition, external interface 762 may be provided as with processor 752, so that mobile computing is set Standby 750 can be with the near region field communication of other equipment.For example, external interface 762 can provide in some embodiments it is wired Communication provides wireless communication in other embodiments, and multiple interfaces can also be used.

Memory 764 stores the information in mobile computing device 750.Memory 764 can be implemented as computer-readable One in medium or medium, one or more volatile memory-elements or one or more Nonvolatile memery units or It is multiple.Extended menory 774 may be provided with and be connected to mobile computing device 750, example by expansion interface 772 Such as, the expansion interface may include single in-line memory modules (SIMM) card interface.Such extended menory 774 can be to set Standby 750 provide additional memory space, or can also store the application or other information for mobile computing device 750.Tool Body, extended menory 774 may include the instruction for executing or supplementing process described herein, and can also include Security information.Thus, for example, extended menory 774 may be provided as the security arrangement for mobile computing device 750, And the instruction for allowing the safe handling to mobile computing device 750 can be programmed with.In addition, security application can be via SIMM card is provided together with additional information, and such as identification information is placed on SIMM card in a manner of it can not invade.

For example, extended menory 774 may include flash memory and/or NVRAM memory.In an embodiment In, computer program product is tangibly embodied in the information carrier.Computer program product includes instruction, and described instruction works as quilt One or more methods of such as those described above are executed when execution.Information carrier is computer-readable or machine readable Memory on medium, such as memory 764, extended menory 774 or processor 752 can for example pass through transceiver 768 Or external interface 762 receives.

Mobile computing device 750 can be wirelessly communicated by communication interface 766, and the communication interface can be in necessary feelings It include digital signal processing circuit under condition.Communication interface 766 can be provided in the communication under various modes or agreement, such as GSM Audio call, SMS, EMS or MMS message transmitting, CDMA, TDMA, PDC, WCDMA, CDMA2000 or GPRS and other.In this way Communication can for example be occurred by transceiver 768.In addition, short range communication can such as using bluetooth, WiFi or other in this way Transceiver (not shown) occur.In addition, global positioning system (GPS) transceiver devices 770 can be to mobile computing device 750 Additional navigation and location-related wireless data are provided, it can be in appropriate circumstances by operating in mobile computing device On 750 using.

Mobile computing device 750 can also use audio codec 760 audibly to communicate, and the audio codec can To receive verbal information from the user and the verbal information received be converted into digital information.Audio codec 760 The sound audible to user can be similarly for example generated by the loudspeaker in the telephone receiver of mobile computing device 750 Sound.Such sound may include sounds from voice phone calls, may include record sound (for example, speech message, Music file etc.) and can also include the sound generated by the application run on mobile computing device 750.

As shown in figures, mobile computing device 750 can be implemented in many different forms.For example, mobile meter Calculating equipment 750 may be implemented as cellular phone 780.Mobile computing device 750 can also be implemented as smart phone 782, a A part of personal digital assistant or other similar mobile device.

The various embodiments of system and technology described herein can be implemented in Fundamental Digital Circuit, integrated electricity In road, the ASIC specially designed (specific integrated circuit), computer hardware, firmware, software and/or combination thereof.These various realities The mode of applying can include with the embodiment of one or more computer programs, and one or more of computer programs are to wrap Include can be performed and/or interpretable, institute on the programmable system that can be at least one dedicated or general programmable processor At least one programmable processor is stated to be coupled to connect from storage system, at least one input equipment and at least one output equipment It receives data and instruction and sends data and instruction to storage system, at least one input equipment and at least one output equipment.

These computer programs (also referred to as program, software, software application or code) include being directed to programmable processor Machine instruction, and can come with level process formula and/or Object-Oriented Programming Language and/or with compilation/machine language real It applies.As used in this article, term " machine readable media " and " computer-readable medium " refer to for programmable processor Any device and/or equipment of machine instruction and/or data are provided (for example, disk, CD, memory, programmable logic device (PLD)), the machine readable media including reception as the machine instruction of machine-readable signal.Term " machine-readable signal " is Refer to for providing any signal of machine instruction and/or data to programmable processor.

In order to provide the interaction with user, system and technology described herein can be carried out on computers, described Computer has for showing the display equipment of information (for example, cathode-ray tube (CRT) or liquid crystal display (LCD) prison to user Visual organ) and user can by its to computer provide input keyboard and pointing device (for example, mouse or trackball).Its The equipment of his type also can be used in providing the interaction with user；For example, the feedback for being provided to user can be any form Sense feedback (for example, visual feedback, audio feedback or touch feedback)；And input from the user can be to include the sense of hearing Any form of input, voice input or tactile input is received.

System and technology described herein can be carried out in computing systems, and the computing system includes back-end component (for example, as data server), or including middleware component (for example, application server), or including front end component (for example, Client computer with graphical user interface or web browser, user can be clear by the graphical user interface or network Device of looking at is interacted with the embodiment of system described herein and technology), or including such back-end component, middleware component Or any combination of front end component.The component of the system can be by any form or medium of digital data communications (for example, logical Communication network) it is connected with each other.The example of communication network includes local area network (LAN), wide area network (WAN) and internet.

The foregoing description of embodiment provides diagram and description, but be not intended to be exhaustive or be not intended to and limit embodiment In disclosed precise forms.In view of above disclosure or can be possible to from the practice of embodiment make modification and Modification.

It is evident that exemplary aspect as described above can be in the embodiment that illustrates in attached drawing Software, firmware and hardware many different forms implement.Actual software code or special in terms of for implementing these Control the hardware property of should not be construed as limited to.Therefore, it is described in the case where not quoting specific software code described The operation and behavior of aspect --- it should be appreciated that software and control hardware can be designed to implement based on description herein The aspect.

In addition, certain parts of embodiment may be implemented as executing " component " of one or more functions.The component It may include the combination of the hardware or hardware and software of such as processor, ASIC or FPGA.

Even if the specific combination of feature is documented in claims and/or is disclosed in the description, these combinations It is not intended to the disclosure of limitation possible embodiment.In fact, many features in these features can be with not specific note The mode for being loaded in claims and/or not specifically disclosing in the description is combined.Although each of be listed below from Other claims of only one can be depended directly on by belonging to claim, but the disclosure of possible embodiment includes every The combination for the every other claim that a dependent claims and claim are concentrated.

Element used herein, movement or instruction be not construed as it is crucial or essential, except being far from it It is expressly recited.In addition, as used in this article, word " one " and "one" be intended to include it is one or more, and can be with " one It is a or multiple " it is interchangeably used.When being intended to only one, term "one" or similar language are used.In addition, unless another Have and clearly state, phrase " being based on " is intended to mean " being at least partially based on ".

Claims

1. a kind of method, comprising:

Translation and inquiry is received by the one or more processors of equipment,

The translation and inquiry request is by one or more terms from source language translation at object language；

One or more translation features associated with the translation and inquiry are determined by one or more of processors；

Feature is assigned by one or more of processors come each translation feature into one or more of translation features Value is to form one or more features value；

Feature weight is applied to one or more of spies in a manner of linearly or nonlinearly by one or more of processors Each characteristic value in value indicative is to generate end value；

It is determined and dialogue translation user to be provided based on whether the end value meets threshold value by one or more of processors Interface or non-conversational translate user interface,

The dialogue translation user interface promotes the translation to session,

Non-conversational translation user interface one or more translation searches are provided as a result, and

The non-conversational translation user interface is different from dialogue translation user interface；And

The dialogue, which is provided, by one or more of processors when the end value meets the threshold value translates user interface For showing.

2. according to the method described in claim 1, further include:

The non-conversational translation user interface is provided when the end value is unsatisfactory for the threshold value for showing.

3. according to the method described in claim 1, further include:

Receive dialogue training data and non-conversational training data；

It is trained based on one or more of translation features, the dialogue training data and the non-conversational training data point Class device；

Based on generating multiple feature weights to the training of the classifier,

The feature weight for each characteristic value being applied in one or more of characteristic values is from the multiple feature It is selected in weight；

Housebroken classifier is generated based on the multiple feature weight；

Store the housebroken classifier；And

It is received from the housebroken classifier described in each characteristic value being applied in one or more of characteristic values Feature weight.

4. according to the method described in claim 1, further include:

The translation to the translation and inquiry is generated with associated with the translation and inquiry object language；

It is following for showing to provide to translate user interface via the dialogue:

With the translation to the translation and inquiry of the object language, and

With the translation and inquiry of the original language；

The potential response to the translation and inquiry is generated with associated with the translation and inquiry object language；And

User interface is translated via the dialogue to provide with the potential response of the object language for showing.

5. according to the method described in claim 1, further include:

Receive the instruction that user ratifies the dialogue translation user interface；And

The threshold value is changed based on the instruction.

6. according to the method described in claim 1, further include:

Receive the request for the information from the dialogue translation user interface to be exported to message transmission application；And

The information from the dialogue translation user interface is exported into the message transmission application based on the request.

7. according to the method described in claim 1, further include:

Translate user interface via the dialogue providing for changing the original language associated with the translation and inquiry or The option of the object language；And

Change the original language or the object language when the option is selected by the user.

8. according to the method described in claim 1, further include:

User interface is translated via the dialogue to provide with the original language associated with the translation and inquiry and the mesh The conversation history of poster speech.

9. a kind of system, comprising:

Equipment, the equipment include:

For receiving the device of translation and inquiry,

For determining the device of one or more translation features associated with the translation and inquiry；

Characteristic value is assigned for each translation feature into one or more of translation features to form one or more spies The device of value indicative；

Each characteristic value for being applied to feature weight in a manner of linearly or nonlinearly in one or more of characteristic values To generate the device of end value；

For determining that provide dialogue translation user interface or non-conversational translates based on whether the end value meets threshold value The device of user interface,

The dialogue translation user interface promotes the translation to session,

The non-conversational translation user interface is different from dialogue translation user interface；

For the dialogue translation user interface being provided when the end value meets the threshold value with the device for display；With And

For the non-conversational translation user interface being provided when the end value is unsatisfactory for the threshold value with the dress for display It sets.

10. system according to claim 9, wherein the equipment further include:

For generating the device of the translation to the translation and inquiry with the object language associated with the translation and inquiry；

It provides below for translating user interface via the dialogue with the device for display:

With the translation to the translation and inquiry of the object language, and

With the translation and inquiry of the original language；

For being generated with the object language associated with the translation and inquiry to the potential response of the translation and inquiry Device；And

It provides for translating user interface via the dialogue with the potential response of the object language for showing Device.

11. system according to claim 9, wherein the equipment further include:

For receiving the device of dialogue training data and non-conversational training data；

For being instructed based on one or more of translation features, the dialogue training data and the non-conversational training data Practice the device of classifier；

For based on the device for generating multiple feature weights to the training of the classifier,

For generating the device of housebroken classifier based on the multiple feature weight；

For storing the device of the housebroken classifier；And

For receiving each characteristic value being applied in one or more of characteristic values from the housebroken classifier The device of the feature weight.

12. system according to claim 9, wherein the equipment further include:

For receiving it is used to that the dress of the request of message transmission application will to be exported to from the information of the dialogue translation user interface It sets；And

For the information from the dialogue translation user interface to be exported to the message transmission based on the request The device of application.

13. system according to claim 9, wherein the equipment further include:

The device for ratifying the instruction of the dialogue translation user interface for receiving user；And

For changing the device of the threshold value based on the instruction.

14. system according to claim 9, wherein the equipment further include:

It provides for translating user interface via the dialogue for changing the source language associated with the translation and inquiry The device of the option of speech or the object language；And

For changing the device of the original language or the object language when the option is selected by the user.

15. system according to claim 9, wherein the equipment further include:

It provides for translating user interface via the dialogue with the original language associated with the translation and inquiry and institute State the device of the conversation history of object language.